I don’t know the answer to my own question! However, based on the history of technology development, there is a good chance that Web user interfaces will gradually phase out over the next several years as the widespread use of AI agents takes over the way we interact with websites.
In this article, I look at 40 years of tech history so you can be prepared should AI agents be the way your customers want to interact with your organization’s products and services.
Today, we communicate with many of our personal and enterprise software applications via separate Web user interfaces. I go to the JetBlue or American Airlines website applications to book a plane ticket, Ticketmaster or StubHub to buy concert tickets, QuickBooks for my business accounting. All are accessed via proprietary web-based applications.
The model of proprietary user interfaces for each software application is likely to go away because of Agentic AI. The future is a world where our AI agents will run the software and humans will enter commands into the agentic application rather than into individual user interfaces.
In the future, I may go to my AI agent to create an invoice in QuickBooks or to buy a plane ticket or concert ticket.
This is true of both personal software and business software.
If you’re not already using an AI agent, Meta’s new Muse consumer application released a few weeks ago is a good way to experience the future. Muse connects email, calendar, Instagram and other apps. If a task needs a tool that doesn't exist, Muse builds it for you. Give Muse a goal or a task and it builds an action plan, tracks your progress and proactively takes tasks off your plate. It learns and gets sharper, returning ideas and suggestions built around your needs, ready to act on them under your direction.
I’ve seen the transition from separate applications to a combined application several times in my career. At the dawn of the personal computer era, you could only run one program at a time. If your word processor was running, you couldn’t run a spreadsheet. I recall the day I first saw Microsoft Windows. This would have been about 1987 or so. Wow! I could have multiple apps open at once! Crazy!
To work within MS Windows, app developers needed to make sure their programs could run correctly within the Microsoft ecosystem. Microsoft quickly became the default UI for business software.
Same thing back in 2007 when I got my iPhone 1. I no longer needed to carry a mobile phone, pager, and iPod. They were on the same device!
To work within the Apple mobile ecosystem, organizations needed to create an app and submit it to Apple for approval. With the app store (and a similar ecosystem for Android), your phone became a default UI.
Today there is a raging battle over who will control the Agentic AI world.
The Meta Muse launch clearly signals that Zuck and his team wants to control the UI ecosystem for personal applications.
Meantime, CEO Satya Nadella has said recently that Microsoft’s strategy is for Copilot to be the Operating System for work. A Microsoft blog announcement says: “We’re reimagining Microsoft Copilot to enable work as it evolves and to help expand what every individual and every organization can accomplish in the flow of human ambition. Today we’re introducing the new Copilot to connect the tools people rely on with the next generation of capabilities they’ll need to build, customize and scale AI across work.”
Microsoft and Meta both know that AI agents will likely control software vs. individual user interfaces.
If you believe that propriety user interfaces will become less important, it has powerful ramifications for how to structure work. There are two important aspects that organizations need to grapple with.
If you have a stand-alone web application that clients use, you should be planning for this future. One way is to adopt Model Context Protocol (MCP), an open-source framework for allowing AI agents to talk to software. Many applications already allow access via MCP which allows AI agents to interface with the underlying data, bypassing the native UI.
For example, the QuickBooks MCP Server lets AI assistants access QuickBooks data via a standard interface. It uses the Model Context Protocol to expose QBO features as callable tools, enabling developers to build AI apps that fetch real-time QBO data through MCP.
What Organizations need going forward are people who can understand this new world, learn the Agentic AI tools to interface with business software, and architect how all those components work together.
If you follow this logic, the important thing to think about is: How will Agentic AI coordinating all aspects of a business within several years affect how you create products and work within your organization?